The Forbidden Index: How We Turned OpenSearch Into a Privacy-Preserving Superhero
summary
This project converts OpenSearch into a privacy-aware analytics engine by adding an PII auto-redaction plugin that scans ingest pipelines (regex + NLP), tokenizes sensitive values into a vault-backed mapping, and applies role-based dynamic masking; benchmarking shows ~8 ms ingest overhead and 94% PII detection in tests. Aggregations use differential privacy (example eps = 1 yields different noisy counts per query). The team open-sources a privacy score audit tool that scores indexes 0–100 and lists fixes.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:26Agenda overview
- 2:07Healthcare search privacy risks
- 3:30Why search leaks secrets
- 5:36GDPR HIPAA CCPA explained
- 7:43Four layers of privacy defense
- 8:46Tokenization first layer
- 11:40Dynamic data masking
- 13:34PII ingest plugin
- 16:29Differential privacy importance
- 18:17Demo setup security posture
- 19:43Breach without privacy controls
- 22:13Tokenization in action
- 24:01Role-based masking
- 26:47Differential privacy queries
- 29:38Privacy score audit tool
- 31:28Key takeaways legal obligations
- 32:33Resources Q&A
related talks
13mSecuring the Data Frontier: Real-World Strategies With OpenSearch
16mA Thrilling Journey into Data-Driven Security Automation
18mPrivacy-Preserving Web Search
55mFirewall Reversed: How Open Source Became Asia's Best Defense
30mFuture Paths for the OpenSearch Security Layer
22mEnhancing Privacy through an Interactive On-demand Incremental Information
29mThreat Modeling for OpenSearch: Identifying and Mitigating Security Risks
17mFine-Grained Authorization for Secure RAG With OpenSearch